Is there a recording of it? It sounds very interesting but I missed it.
Sorry if these were already covered, but some pointers for data quality for those who are new:
At the bottom level there is simple validation of each field, like you would treat the fields of a web form. Can I put a string in a number field, a very long string in a string field etc? There are slightly more subtle cases here, where the data type is fairly loose, e.g. int or string, but business-level constraints mean things should be more restricted. E.g. instead of any int, it should be in the range 1-7 because it’s actually a day of the week. This can also move on to simple validation of sets of fields, e.g. 3 int fields that together should form a valid date.
Next is testing relationships between rows. E.g. each row in the order item table should point to exactly one valid row in the customer and product tables. Or each row in the product category table should point to 0 or 1 other rows in the product category table (to show parent/child relationships between categories).
Beyond that are more subjective or context-dependent tests. The tests above are simple pass/fail kinds of test. However there can be other kinds of test where you can try to define what’s a normal range of something (a value in a particular field, the number of rows that meet some criteria etc), and you can see if you see something outside that range. For instance, normally orders in an online shop are for fewer than 10,000 items at a time, or have a total value of less than £20,000. What happens if you see an order with 50,000 items or for £80,000? Or, normally you receive 10-100 deliveries per week from a supplier, but this week there are 2 or 500. Is that OK?
?).